Enhancing the detection ability of control charts in profile monitoring by adding RBF ensemble model

Abstract

While numerous contributions and applications have been extended in profile monitoring, little attention has been paid to employing machine learning techniques in development of control charts. In this paper, a novel control chart based on artificial neural network is proposed to improve the performance of monitoring general linear profiles in Phase II. Specifically, an ensemble of radial basis functions (RBF) is added to the predefined base control chart to enhance the detection ability of the control chart for monitoring linear profile parameters based on the average run length (ARL) criterion. The performance of the proposed method is evaluated by adjusting the multivariate exponentially weighted moving average (MEWMA) control chart as a base control chart under simple and multiple linear profiles. The simulation results demonstrate that the proposed approach is very efficient than competing existing methods for monitoring linear profile parameters. Moreover, profile diagnosis actions, referring to the identification of shifted parameters, are provided based on the RBF networks. Finally, we provide an example from thermal management to illustrate the implementation of the proposed monitoring scheme and diagnostic method.